Wavelet Packet Decomposition and Neural Network Based Fault Diagnosis for Elevator Excessive Vibration

Abstract

As key elements widely used in buildings, the safety of elevators is closely related to the safety of people. Therefore, effective fault diagnosis in particular is important for elevator. Though there are many researches on diagnosing some key components of the elevator system, most researchers focus on process variables with low frequency and few researchers pay attention to the vibration signals of the elevator. However, it is recognized that the vibration signals contain rich information about the running condition of the elevator system and may be useful for diagnosing the excessive vibration problem. Therefore, this paper presents a fault diagnosis method for excessive vibration of elevator based on three-order wavelet packet decomposition and neural network. First, considering the vibration signals are nonstationary and full of noise, features of both time domain and frequency domain which can represent the vibration signal are extracted by wavelet packet decomposition. Then, a neural network model is built to make binary classifications using normal samples and fault samples based on the extracted features. Case study on a real elevator system illustrates the validity of the proposed method.